57c15b815d
Remove device mapping from shareWithAll memory
...
When shareWithAll memory (e.g., host memory) is allocated, set appId
in hc::AmPointerInfo to -1 to indicate that this memory is not mapped
to any device. Peer checking in ihipStream_t::canSeeMemory is not
necessary if memory is shared with all devices. Thus, it is skipped.
Note that earlier host memory is always mapped to device 0 and HIP
always performs peer checking for all kinds of hipMemcpy. Since the
peer checking process requires context locking, hipMemcpy from/to host
memory always grabs device 0's context lock. Therefore, if there is
another thread holding the context lock of device 0 (e.g.,
hipDeviceSynchronize on device 0), hipMemcpy will have to wait for the
lock until it can actually perform memcpy. This can significantly
deteriorate execution performance.
Signed-off-by: Sarunya Pumma <sarunya.pumma@amd.com >
[ROCm/clr commit: 84aadb9274
2018-07-28 23:15:16 -07:00
a796eedde9
Revert "Use memcpy kernel for all pinned memory cases in hipMemcpy2DAsync"
...
[ROCm/clr commit: f554e48db3
2018-07-02 14:32:11 +05:30
1c2acf1467
TEMP- fix memcpy2dAsync for trsm issue
...
[ROCm/clr commit: 007e2a4b5f
2018-06-15 16:08:29 +05:30
a41cb78337
Fix stream resolution in memcpy2dasync
...
[ROCm/clr commit: 2ae3be9773
2018-06-14 11:58:56 +05:30
dfe8629088
Fix retrieved locked ptr offset
...
[ROCm/clr commit: 68554e155b
2018-06-13 23:10:05 +05:30
e34247ca19
Merge pull request #497 from gargrahul/fix_memcpy3d_fastpath
...
Fix hipMemcpy3D for fast path
[ROCm/clr commit: 53037472ff
2018-06-06 14:44:02 +05:30
914a0aaf9c
Fix hipMemcpy3D for fast path
...
[ROCm/clr commit: 163d4a5b03
2018-06-05 18:54:33 +05:30
3a6e8aadbd
Fix depth value for 3D allocations
...
[ROCm/clr commit: fa6ce7a724
2018-06-04 18:00:22 +05:30
84cb355df6
Fix memcpy2D for malloc+ hostRegister
...
[ROCm/clr commit: a3609eaf61
2018-05-31 13:14:27 +05:30
0ab5c9d6f1
Use 64x4 grid dims
...
[ROCm/clr commit: 27e0566c3a
2018-05-24 23:51:52 +05:30
439655d213
Clean up and fix remaining bytes copy
...
[ROCm/clr commit: aed2653857
2018-05-24 23:30:27 +05:30
4fc64ff4c5
Fix memcpy2d kernel dims
...
[ROCm/clr commit: 53a7c61e9b
2018-05-24 17:00:12 +05:30
4e52a634a1
Correct remaining bytes in copy 2d kernel
...
[ROCm/clr commit: 96b4618d26
2018-05-24 08:27:24 +05:30
432467d88b
Optimize memcpy2D kernel use
...
[ROCm/clr commit: 2c3d1498d4
2018-05-23 14:43:47 +05:30
335dad2feb
Merge pull request #464 from gargrahul/fix_memcpy2d_pinned_mem_case
...
Fixed memcpy2D for pinned memory case using 2D kernel
[ROCm/clr commit: 7042fe6067
2018-05-22 10:42:28 +05:30
54497d1e2b
Fixed memcpy2D for pinned memory case using 2D kernel
...
[ROCm/clr commit: 40fb44dbe6
2018-05-21 22:14:45 +05:30
efe6b5e4a0
hipMemcpy returns success if sizeBytes is 0.
...
Fixes SWDEV-153754 & SWDEV-154178.
[ROCm/clr commit: 66d05e6fc3
2018-05-21 15:38:44 +05:30
8575112052
Fix for memcpy2DAsync for pinned host memory case
...
[ROCm/clr commit: 4f5bdb071c
2018-05-18 21:09:50 +05:30
106f3f591e
Merge pull request #433 from gargrahul/add_hipmemset3d
...
Added hipMemset3D
[ROCm/clr commit: 1c93e11cdf
2018-05-18 14:54:15 +05:30
84089036b0
Fixed hipMemcpy2D to handle 1D memcpy case
...
[ROCm/clr commit: 8413fb51e1
2018-05-16 11:07:10 +05:30
a7be5be448
Added hipMemset3D
...
[ROCm/clr commit: 78568435da
2018-05-07 10:24:30 +05:30
6278830c1a
Null checks added for hipmallocpitch and hipmemcpy apis
...
[ROCm/clr commit: 12d8a47c0c
2018-05-03 09:27:50 +05:30
6f05c37cde
Fix texture 3D for HIP/NVCC
...
[ROCm/clr commit: 1d76c48e3d
2018-05-02 11:56:37 +05:30
06dd83f7f1
SWDEV-141024
...
[ROCm/clr commit: 74faa61d52
2018-04-20 17:40:00 +05:30
39a19cc9e8
Added hipMemset2DAsync support
...
[ROCm/clr commit: fcc0866681
2018-04-17 18:27:27 +05:30
cf6322da63
Correct missed ihipMemsetCopyDataType change
...
[ROCm/clr commit: 2ee80397f2
2018-04-12 10:27:19 +05:30
312b764721
Changed ihipMemsetCopyDataType to ihipMemsetDataType
...
[ROCm/clr commit: 2658963f5b
2018-04-12 09:29:22 +05:30
aff8d83cdf
Fix hipMemset stream resolution
...
[ROCm/clr commit: 88073a17c1
2018-04-11 19:01:53 +05:30
a3bc29976b
hipMemset refactoring
...
[ROCm/clr commit: 40846f6f8e
2018-04-11 15:58:48 +05:30
cdb7f8f354
hipMemcpyAsync returns success when trying to copy 0 bytes
...
Change-Id: I4c0ee7ccc7563e2df657b50356cdd7fec9a1ef15
[ROCm/clr commit: 445de66560
2018-04-09 12:39:44 +05:30
46ddefedee
Apply .clangformat to all repo source files
...
Change-Id: I7e79c6058f0303f9a98911e3b7dd2e8596079344
[ROCm/clr commit: 9e47fccc89
2018-03-12 11:29:03 +05:30
c4d704b2ad
Change directory name to match HIP lowercase style.
...
[ROCm/clr commit: 696f0595a3
2018-02-22 13:15:10 +00:00
2dc84aeaf9
Merge pull request #321 from gargrahul/hipMemcpyArray_Functions
...
Added support for hipMemcpy Array functions-
[ROCm/clr commit: 647d1ba310
2018-02-12 10:36:38 +05:30
f15543d5cf
Fixed host allocated globals address lookup for host usage
...
Fixed texture driver APIs failure
[ROCm/clr commit: b8c23f979b
2018-01-30 18:06:31 +05:30
ec4cc27986
Added support for -
...
- hipMemcpyFromArray
- hipMemcpyAtoH
- hipMemcpyHtoA
[ROCm/clr commit: ca5bcb5af4
2018-01-16 11:44:19 +05:30
467b908368
Added support for
...
- 3D texture driver APIs
- hipMalloc3D
- hipMemcpy3D for destination other than array
[ROCm/clr commit: 105df94cd0
2017-12-05 14:11:13 +05:30
cabf4d817c
Fix some cppcheck style issues.
...
[ROCm/clr commit: 5933ca7300
2017-12-01 20:45:34 +00:00
b1aee08dc1
Fix warning from default cppchek.
...
[ROCm/clr commit: 743c54a012
2017-12-01 20:45:33 +00:00
76336d46b4
Fix legacy mode detection of the address of an agent allocated variable. In this mode, there exist two executables per each code object, one created by HCC and one created by HIP. Since we dispatch through HCC in legacy mode, we should obtain the address for an agent allocated variable from the latter's executable. Also add two omitted validity checks, whose absence could lead to segfaults when the current process had no .kernel section and / or when an invalid or empty blob was extracted from the latter.
...
[ROCm/clr commit: 33bb425013
2017-11-30 03:29:04 +00:00
a27fa25d76
Revert "Revert adoption of CUDA indexing in general - this can only work with later versions of the compiler, just like module based dispatch, and thus must be guarded against usage in earlier (e.g. 1.6) versions."
...
This reverts commit 1cd64164966518846
2017-11-29 21:49:10 +00:00
a1d5f0ab0c
Revert "Revert adoption of CUDA indexing in general - this can only work with later versions of the compiler, just like module based dispatch, and thus must be guarded against usage in earlier (e.g. 1.6) versions."
...
This reverts commit d2fd1f52557000b56
2017-11-29 21:36:29 +00:00
abf92dd6e7
Change memset kernel to use memcpy instead of placement new. Simplify indexers.
...
[ROCm/clr commit: d5c8de3f41
2017-11-28 19:45:47 +00:00
579a3187da
Merge remote-tracking branch 'origin/master' into feature_use_module_based_dispatch_instead_of_pfe
...
# Conflicts:
# src/hip_module.cpp
[ROCm/clr commit: d37a5a6008
2017-11-28 17:29:11 +00:00
212afdf56b
Fixed review comments
...
[ROCm/clr commit: 38029f2849
2017-11-21 21:19:06 +05:30
dc2c56c43d
Changed function hipMemcpy_2D to hipMemcpyParam2D
...
[ROCm/clr commit: 24307fe5c4
2017-11-21 12:36:24 +05:30
1366b2d5ae
Refactor the __device__ versions of memset and memcpy to be less awkward i.e. not return nullptr as opposed to the destination pointer (it can only be assumed it was done for maximum confusion) and actually unroll as they claim to. Change all of the {to, from}Symbol functions to use hipModuleGetGlobal, as opposed to hc::accelerator::get_symbol_address which is no longer valid with module based dispatch.
...
[ROCm/clr commit: f8c1c1b38e
2017-11-21 02:40:34 +00:00
bb8c31c198
Texture driver APIs support
...
[ROCm/clr commit: 1851c153f6
2017-11-09 22:10:55 +05:30
142213a52b
Clean up trailing whitespace so as to reduce noise in #246 .
...
[ROCm/clr commit: 0ad1308efd
2017-11-08 00:08:55 +00:00
21d1d3f1ed
Merge remote-tracking branch 'origin/master' into feature_use_module_based_dispatch_instead_of_pfe
...
# Conflicts:
# src/hip_module.cpp
[ROCm/clr commit: 17753cbd92
2017-11-03 10:53:39 +00:00
840895525a
Correctly deal with functions from shared objects, wherein the program visible VA == so_base_va + st_value(function_symbol). Remove quaint usage of pfe for hipMemset (which is actually fill_n).
...
[ROCm/clr commit: 4693c5e56c
2017-11-01 22:33:13 +00:00
Sarunya Pumma
Rahul Garg
Rahul Garg
Maneesh Gupta
Lakhan Singh
Alex Voicu
Ben Sander